Vincent Wang 4df356d1c9 Train BERT Using BFloat16 on A100 (#6090) ... * traing bert using bf16 * Adam support bf16 * bugfix * add fusedmatmul support * fix after merge from master. * bugfix * bugfix after merge from master * fast reduction for bf16. * resolve comments * fix win build * bugfix * change header file. Co-authored-by: Vincent Wang <weicwang@microsoft.com>		2021-01-14 19:04:32 +08:00
..
bias_softmax.cc	Fuse softmax(a + b) in case of simple broadcast (#4937)	2020-09-18 14:15:55 -07:00
bias_softmax.h	Reduce amount of compiled CUDA device code (#6118)	2020-12-14 15:27:40 -08:00
bias_softmax_impl.cu	GatherGrad optimization (#5524)	2020-10-26 12:53:53 -07:00
binary_elementwise_ops.cc	Add GPU kernels for ROCm EP (#5655)	2020-11-06 16:11:06 -08:00
binary_elementwise_ops.h
binary_elementwise_ops_impl.cu	Minor changes to AMD element-wise kernels to converge with CUDA element-wise kernels.	2020-12-15 08:46:36 -08:00
binary_elementwise_ops_impl.h
complex_mul.cc
complex_mul.h
complex_mul_impl.cu
complex_mul_impl.h
cufft_plan_cache.h
fft_ops.cc
fft_ops.h	Reduce amount of compiled CUDA device code (#6118)	2020-12-14 15:27:40 -08:00
fft_ops_impl.cu
fft_ops_impl.h
fused_matmul.cc	Train BERT Using BFloat16 on A100 (#6090)	2021-01-14 19:04:32 +08:00